create-eval

Generate promptfoo YAML, bash test scripts, or VHS .tape files from conversations.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/troylar/anteroom --skill create-eval-troylar
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: create-eval
Source: https://github.com/troylar/anteroom/tree/main/src/anteroom/cli/default_skills/create-eval
Command: npx skills add https://github.com/troylar/anteroom --skill create-eval-troylar

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Turn informal descriptions or chat conversations into repeatable, deterministic tests and demo recordings so teams can verify AI behavior and reproduce interactions reliably.

Core Features & Use Cases

  • Generate promptfoo evals from a natural-language description or extracted conversation prompts, using an exec provider that runs aroom exec --json for deterministic checks.
  • Create shell test scripts that run aroom exec with deterministic flags and perform structural assertions on exit codes and tool usage.
  • Produce VHS demo tapes that replay conversation prompts as terminal recordings for demos or onboarding.
  • Use Case: Convert a problematic conversation into a promptfoo YAML that asserts the assistant used read_file before editing files and then run it in CI to catch regressions.

Quick Start

Create a promptfoo eval that verifies the assistant used read_file when summarizing the project's README.

Frequently Asked Questions about create-eval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I convert chat transcripts into promptfoo evals?

To convert chat transcripts into promptfoo evals, you can generate promptfoo YAML configurations from natural-language descriptions or existing conversation logs, using an exec provider that runs aroom exec --json for deterministic behavioral checks.

Can I create deterministic shell test scripts from natural language descriptions?

Yes, you can create deterministic shell test scripts from natural language by generating bash scripts that run aroom exec with deterministic flags and perform structural assertions on exit codes and tool usage.

What is the best way to record terminal demos from conversation prompts?

The best way to record terminal demos from conversation prompts is by producing VHS .tape files that replay the interactions as terminal recordings for onboarding or demonstration purposes.

How do I verify if an AI assistant used a specific tool like read_file before editing?

To verify if an AI assistant used read_file before editing, you can convert the problematic conversation into a promptfoo YAML that includes structural assertions validating tool usage, and run it in CI to catch regressions.

Does this approach work for catching AI regressions in CI pipelines?

Yes, this approach works for catching AI regressions in CI pipelines by converting extracted conversation prompts into deterministic regression tests and reproducible terminal demos for local and CI verification.